基于偏最小二乘回归的城镇登记失业率预测

选取了城镇登记失业率的五类影响因素和18个预测指标，采用本征值与病态指数检验法进行多重共线性进行诊断，说明在18个预测指标之间存在着严重的多重共线性，为了克服多重共线性导致的预测失真问题，运用偏最小二乘回归模型，以我国城镇登记失业率为研究对象，按照交叉有效性的决策原则，建立了预测模型，所提取成分对x累计解释能力达到99．6％，对Y的累计解释能力达到96~／oo，具有较高的预测精度，该方法还可进一步应用到类似问题中。

Urban Registered Unemployment Rate Forecasting Base on Partial Least-Squares Regression

Firstly, the urban registered unemployment rate in the 5 kind of influence factors and the 18 predictors are determined, then the ei- genvalue and the morbid state index inspection method are used to carry on the diagnosis of multicollinearity, the results show that exists severe multi- collinearity between 18 predictors, Secondly in order to overcome the prediction distortion caused by multicollinearity problems, the partial least-squares model is applied to forecasting the urban registered unemployment rate of China, the model has a higher forecast accuracy, its cumulative interpretation ability of X and y are 99.60/00 and 96~ , the method can also be further applied to similar problems,